Puccinia striiformis urediniospores were obtained from the susceptible wheat cultivar Mingxian 169 in an experimental field (approximately 3 m × 6 m area) at the Northwest A&F University, Yangling, Shaanxi, China in the summer of 2015. Samples were collected and stored in a desiccator at 4°C. The Puccinia striiformis species identity of these field samples were validated based on the complete internal transcribed spacer (ITS) sequence of ribosomal DNA (rDNA). Total genomic DNA was extracted from uredinospores of P. striiformis with CTAB method as described by Justesen et al. (2002). The ribosomal rDNA-ITS primers, ITS1 (5′-TCCGTAGGTGAACCTGCGG-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′), were synthesized by Sangon Bio-Tech Co., Ltd. Polymerase chain reaction (PCR) amplification was done following standard methods according to Zheng et al. (2017).

DsRNA was extracted from 1.0 g of P. striiformis urediniospores according a described method with minor modifications (Zheng et al., 2014), and absorption column made up of cellulose powder CF-11 (Whatman, United Kingdom) was used. Extracted dsRNAs were treated with DNase I and S1 nuclease (TaKaRa Bio Inc) to remove genomic DNA and single-stranded RNA (ssRNA) contaminations, the qualities of which were then analyzed based on 1.0% (w/v) agarose gel electrophoresis.

The dsRNA sample (1.0 μg) was used for cDNA library construction using the NEBNext^®Ultra^TM RNA Library Prep Kit (Illumina, United States) following manufacturer’s instructions. The cDNA is end-repaired and adenylated prior to adaptor ligation, library construction and amplification under this method. Then the sequenced-ready library was subjected to 5 million of 150 nucleotide (nt) paired-end reads using Illumina HiSeq 4000 technology. The cDNA library construction and deep sequencing analysis were carried out by Shanghai Hanyu Bio-Tech Co., Ltd. After deep sequencing (>40,000 × coverage), raw reads were cleaned by removing adapter sequences and low-quality bases (PHRED quality scores ≤ 5), and truncated reads smaller than 35 bp were discard. 29 contigs were obtained and the N50 length is 1,489 nt. The total length of sequencing is about 37 kb with a GC content of 42.06%. De novo assembly of contiguous sequences was conducted using the Velvet de novo assembly algorithm with k-mer: 69. Minimum contig length was 500 bp as well as minimum coverage was 18. The ends and the other parts of the sequences were all confirmed by Sanger sequencing. To obtain the termini of the dsRNAs, rapid amplification of cDNA ends (RACE) was performed (Zheng et al., 2013).

Open reading frames (ORFs) were identified using the National Center for Biotechnology Information (NCBI) ORF Finder program¹. Motif searches were performed in three databases, including PROSITE², Pfam³ and CDD⁴. RNA pseudoknot structure was predicted using the DotKnot program and drawn with PseudoViewer3 (Byun and Han, 2009; Sperschneider and Datta, 2010).

Phylogenetic trees were constructed based on the deduced amino acid sequences of the putative RdRp and CP regions using the maximum-likelihood (ML) method of the MEGA program (version 6.0)with 1,000 bootstrap replicates as described previously with minor modifications (Kondo et al., 2016). Multiple alignments of the sequences of RdRp and CP were conducted with Clustal-X program (Thompson et al., 1997).

Virus particles were purified using the method as described previously with minor modifications (Sanderlin and Ghabrial, 1978). Three gram P. striiformis urediniospores were grounded into fine powder in the presence of liquid nitrogen with a sterilized mortar and pestle. The powder was then mixed with 200 ml extraction buffer made up of 0.1 M sodium phosphate, and pH 7.0 containing 3% Triton X-100. The suspension was centrifuged at 10,000 × g for 20 min to remove the spore cell debris. Subsequently, the supernatant was subject to a 1.5 h ultracentrifugation at 120,000 × g for viral particle precipitation. The pellets were then suspended in 0.1 M sodium phosphate buffer and centrifuged at 15,000 × g for 30 min. Then the suspension containing the virus particles was fractionated via a 10–40% (w/v) sucrose gradient by centrifugation at 60,000 × g for 3.5 h. Fractions in the middle portion of the tube were collected, and stained with 2% phosphotungstic acid (pH 7.4) and observed under a transmission electron microscope (TEM) (HT7700, Japan).

The dsRNAs from virus particles was extracted using phenol, chloroform and isoamyl alcohol, and then subjected to electrophoresis in 1% (w/v) agarose gel. Using the dsRNAs from virus particles as templates, complementary DNAs were synthesized as described by Xie et al. (2011) with tagged random dN6 primers (Rong et al., 2002; Zheng et al., 2013). Reverse transcription polymerase chain reaction (RT-PCR) was then performed with specific primers (Supplementary Table S1) which target the four PsVs respectively according to the method of Vainio et al. (2012) with minor modifications.

To detect dsRNA virus, P. striiformis urediniospores were collected from heavily infected wheat leaves growing in the field (Figure 1A). Urediniospores were round-shaped and approximately 30 μm in diameter (Figure 1B). The sequence length of ITS is 663 bp and it has 99% sequence identity to the P. striiformis strain CYR32. The result confirmed the fungus belonged to P. striiformis species. Nucleic acid extraction obtained nucleic acid bands of approximately 5.0 kb in size (Figure 1C) that resisted DNase I and S1 nuclease digestions, indicating the presence of dsRNA-like nucleotides.

Four P. striiformis-derived totivirus-like sequences, termed as Puccinia striiformis virus 1 to 4 (PsV1 to 4) (Supplementary Table S2), were identified from the post-assembly contigs by deep sequencing. The 5′- and 3′- untranslated regions (UTRs) of all PsVs were obtained. The complete nucleotide sequences of PsVs (1 to 4), ranging from 5,008 to 5,061 nt, were all deposited in GenBank under accession numbers KY207361-KY207364, respectively.

Pair-wise comparisons among proteins encoded by PsV1 to 4 revealed moderate levels (37–42%) of amino acid sequence identity of RdRp, but low-level similarities (31–33%) among CP. It is notable that the 5′-end sequences of PsVs 1 to 4 (5′-AUAAAUCCCC…-3′, 5′-AUACAAUCCCC…-3′, 5′-AUAUAACUCCC…-3′ and 5′-AUAAAACCCCC…-3′, respectively) appeared to be partially conserved. Likewise, the 3′-end sequences were not conserved (data not shown). All four genomes contained two ORFs encoding CP and RdRp respectively (ORF1 and ORF2) (Figure 2), such organization is typical of totiviruses. The predicted CP and RdRp proteins showed moderate levels of amino acid sequence identities (32–34% for CP and 39–41% for RdRp, respectively) with Phakopsora pachyrhizi mycovirus (PpV) and red clover powdery mildew-associated totivirus 5 (RPaTV5) (Supplementary Table S3). A search of conserved domain database (CDD) and multiple protein alignment confirmed that the predicted RdRp domains contain eight conserved motifs (I to VIII), including the GDD motif, which are the typical characteristics of mycovirual RdRps (Figure 3) (Routhier and Bruenn, 1998).

The PsV sequences between position -3 and +4 relative to the ORF2 AUG start codons (Figure 4A) are similar to the translation initiation sites of the viruses of plants and yeasts (Lütcke et al., 1987; Hamilton et al., 1987). Sequence analysis of PsVs indicated that there is an overlap region between ORFs 1 and 2 (Figure 2). It is therefore possible that ORFs 2 of PsVs is translated as a fusion protein with ORFs 1 through a -1 ribosomal frameshift which is a canonical slippery sites ‘XXXYYYZ’ (where X is A/C/G/U, Y is A/U, and Z is A/C/U) within the overlapping region (Bekaert and Rousset, 2005). The slippery site ‘GGA/GUUU’ sequence in PsV1 to 4 sequences (Figure 4B) is similar to these of the RPaTVs (Kondo et al., 2016). Using the DotKnot program, a pseudoknot structure was predicted in the downstream of each putative slippery site (Figure 4B). RNA pseudoknots are known to help pausing the translating ribosome and increasing the frequency of frameshifting (Plant et al., 2003; Zhai et al., 2008; Kondo et al., 2016).

The maximum likelihood (ML) tree for RdRP is shown in Figure 5, Totivirus contained four subgroups, I-A, I-B, I-C and I-D. PsV1 to 4 clustered with RPaTVs 5 to 8 and PpV in the subgroup I-D. The ML phylogeny based on CPs is shown in Figure 6, which had similar topology as the one based on RdRPs.

Taken together, genome organizations, amino acid sequence alignments and phylogenetic analyses all support that PsV1 to 4 are new members of the genus Totivirus within the family Totiviridae.

Under TEM, the viral particles purified from P. striiformis urediniospores were isometric with an average diameter of 35 nm (Figure 7A). The sizes of dsRNAs from viral particles were similar with the dsRNAs extracted directly from P. striiformis urediniospores (Figure 7B). Using the dsRNAs extracted from viral particles as templates, we performed RT-PCR based on PsV-specific primers targeting RdRps, which successfully amplified products and in all cases the sizes were identical to ones expected based on the PsV genomes (Figure 7C).